Thomson Reuters flags $143B risk from AI execution gap

This report is a reminder that raw model capability isn’t the same as organizational transformation. In law, tax, audit and risk—fields that will be early heavy users of AGI‑class systems—firms have rushed to give staff AI tools but haven’t built the workflows, governance and training that actually turn those tools into better outcomes. The result is shadow AI, compliance headaches and frustrated professionals who are ready to walk if their employers don’t get serious.

Why it matters for the AGI race is that adoption will be lumpy. Even if near‑AGI models become widely available, many high‑stakes sectors will under‑realize the potential because they can’t integrate AI into regulated processes without breaking trust or the law. That creates an opening for a smaller number of firms that do crack the execution puzzle to pull away dramatically, compounding both economic and knowledge advantages as they embed AI deeper into their workflows.

It also underscores a subtle safety dynamic: when sanctioned tools are clunky, people route around them. The more organizations rely on unsanctioned consumer AI, the harder it is to enforce guardrails and audit trails. That suggests that building secure, explainable, enterprise‑grade AI isn’t just a compliance box—it’s a safety strategy that channels usage into systems you can monitor and improve.